by Sydney -Defeated generative AI models from Google and Openai in the international mathematics competition, but the programs reached the points at the gold level for the first time, and the speed with which they improve can cause human self-observation.
In contrast to five young people, none of the AI models achieved the full grades at the international mathematical Olympics (IMO), a prestigious annual competition in which the participants must be under 20 years old.
Google said on Monday that an advanced version of his Gemini Chatbot had solved five of the six mathematical problems that took place in Australia’s Queensland this month.
“We can confirm that Google Deepmind has reached the much -soiled milestone and scored 35 out of 42 points -a gold medal evaluation,” the US Tech giant, the IMO President Gregor Dolinar quoted. “Your solutions were surprising in many ways. Imo -Grader found them clear, precisely and most of them.”
Around 10% of the human participants won medals at the gold level and five received perfect points of 42 points.
Us Chatgpt Maker Openai said that his experimental argumentation model also achieved 35 points on the gold level on the test.
The result “has achieved a long challenge in the KI” for “the most respected math competition in the world,” said Openai researcher Alexander Wei in a social media post.
“We have evaluated our models for the IMO problems of 2025 under the same rules as human participants,” he said. “For each problem, three former IMO medalists have evaluated the submission of the model independently.”
Google achieved a silver medal evaluation at the IMO last year in the city of Bath in the southwest of England and solved four of the six problems.
This lasted two to three arithmetic for longer than this year when his Gemini model solved the problems within the 4.5-hour period.
The IMO said that technology companies had “privately tested AI models for closed sources on this year’s problems”, the same, to whom 641 competing students from 112 countries were confronted.
“It is very exciting to see progress in the mathematical skills of AI models,” said IMO President Dolinar.
The organizers of the competition could not check how much computing power was used by the AI models or whether human participation was given.
In Interview with CBS ’60 minutes At the beginning of this year, one of Google’s leading AI researchers predicted that computers with cognitive skills would be made on a human level within just five to ten years-a landmark that is known as “artificial general intelligence”.
Google Deepmind -CEO Demis Hassabis predicted This AI technology was on the right track to understand the world in a nuanced way and not only solve important problems, but also to develop a feeling of imagination within a decade, thanks to an increase in investments.
“It moves incredibly quickly,” said Hassabis. “I think we have an exponential improvement curve. Of course, the success of the field in recent years has even more attention, more resources, more talents. This increases this exponential progress.”
Detroit Lawnmower Gang is still strong after 15 years
The chairman of the Fed Jerome Powell talks about large banking conferences in the middle of Trump criticism
The White House removes WSJ reporter from Trump’s Scottish journey through paper’s Epstein history
